Elsevier launches SciBite Chat: "ChatGPT for scientists"

Published: 13-May-2024

The academic publishing company has launched SciBite Chat with a database that is underpinned by deep domain expertise, and more than 20 million expertly created scientific terms and their synonyms

You need to be a subscriber to read this article.
Click here to find out more.

Elsevier, an academic publishing company specialising in scientific, technical, and medical content, has launched SciBite Chat.

SciBite Chat is a new AI-powered LLM conversational tool that is informed by SciBite Search’s existing database.

The database is underpinned by deep domain expertise, and more than 20 million expertly created scientific terms and their synonyms.

Elsevier created SciBite Chat as Gen AI is being used to progress innovation in many sectors and the growing use of public LLMs including ChatGPT.

However, “tools like ChatGPT hallucinate answers and references unsuitable for R&D,” Elsevier stated.

SciBite Chat is a new AI-powered LLM conversational tool

SciBite Chat leverages Elsevier’s data technology expertise, combining two powerful technologies:



 

  • Semantic search for accurate and traceable information retrieval
  • LLMs for interpreting human language and answer generation 

By combining these technologies, SciBite Chat has the potential to transform the search experience, providing precise and traceable results while understanding and summarising human language effortlessly.

Additionally, To ensure trust and replicability, SciBite chat provides links to references and ontologies with each answer.

Researchers can also ask questions and receive answers in their native language.

Unlike conventional search tools, SciBite Chat’s natural language query and iterative chat features allow users to have a conversation with their data. 

Tools like ChatGPT hallucinate answers and references unsuitable for R&D

Trust and traceability are also at the core of the user experience with verbatim evidence highlighting and underlying query can be used to identify relevant documents. 

Further, users can switch to the equivalent SciBite Search query language, which helps with explainability and reproducibility.

In the world of life science, where innovation depends on evidence-based insights, SciBite Chat bridges the gap between commoditised LLMs and domain expertise, enabling data democratisation

The three pillars of SciBite Chat 

SciBite Chat is built atop three pillars of data-driven insights:

Accuracy- Accuracy is enhanced by SciBite’s deep expertise in life sciences 

Transparency- Transparency is guaranteed through human explainability incorporated into every step of the user journey. 

Flexibility- Flexibility is in incorporating internal terminologies, data ingestion, and deployment options ensures that organisations can easily extend SciBite Chat to their own world. 

SciBite Chat is poised to become an essential tool for both researchers and organisations, transforming the way they interact with and utilise data.

Elsevier's latest development 

SciBite Chat is the latest development from Elsevier, following the launch of the GAI-powered ScopusAI earlier this year. 

Neal Dunkinson, VP of Solutions and Professional Services at SciBite, Elsevier, said: “SciBite Chat’s use of ontology-backed semantics with Retrieval Augmented Generation (RAG) architecture offers multiple benefits to researchers in life science-related industries.”

SciBite Chat bridges the gap between commoditised LLMs and domain expertise

“The approach improves search results by using the domain expert knowledge captured in ontologies to provide the most relevant documents, grounding Generative AI,” Dunkinson concluded.

You may also like